It looks like the o1-preview model is not working as intended. An error message pops up every time I try to send a message–sometimes I have to send it 8-10 times to even get a response. I contacted support and they were pretty much useless and told me “it just takes longer” when I’ve already explained it’s returning an error message to contact Julius. That was 30 days ago.
Even when you remove any of Julius’ wrappings, such that you’re dealing directly with only with the model, the inference context that you are allowed to input is smaller than even OpenAI allows. So, it’s basically just for show and for sure false advertising.